Contextual marketing system based on predictive modeling of users of a system and/or service
Predictive modeling within a special purpose hardware platform to determine scenarios that are most likely to increase conversion potential for each trial user and retention potential for each active subscriber of a service, collectively referred to as a propensity score. The predictive models are integrated with a contextual marketing system that uses a loss risk assessment to learn user behavior and optimize content messaging designed to improve actual conversion or retention behavior for the user.
1. A computer implemented method for determining churn risk associated with a user of a trial version of a software service and generating messaging content to minimize the risk, said method being performed on a computing device and executed by a processor, said method comprising:
accessing data of the user at a data source, the data of the user comprising historical data and current state data associated with the user;
parsing the accessed data of the user to identify event instances;
determining an initial propensity score based on the historical data and the current state data associated with the user;
in response to determining that the initial propensity score indicates a high churn risk associated with the user:
identifying a first hypothetical scenario associated with a first trained hypothetical model of a first machine learning model and a second hypothetical scenario associated with a second trained hypothetical model of a second machine learning model, wherein each of the first and the second hypothetical scenarios include a corresponding associated software service provider feature, among a plurality of software service provider features, a corresponding hypothetical attribute of the user, and a corresponding hypothetical state data of the user, wherein the first machine learning model and the second machine learning model were trained by segmenting a collection of historical data and current state data, stored at the data source as events and state vectors, into a training set, a test set, and a validation set and passing the training set, the test set, and the validation set through the first machine learning model and the second machine learning model;
implementing the first hypothetical model of the first machine learning model to determine a first propensity score associated with the first hypothetical scenario based on the corresponding hypothetical attribute and the corresponding hypothetical state data, wherein the first propensity score represents a higher propensity score gain;
implementing the second hypothetical model of the second machine learning model to determine a second propensity score associated with the second hypothetical scenario based on the corresponding hypothetical attribute and the corresponding hypothetical state data, wherein the second propensity score represents a lower propensity score gain;
determining whether the first or the second propensity score meets a predetermined threshold;
generating message content to deliver to the user based on the determined propensity score that exceeds the predetermined threshold and at least one service provider feature; and
updating the state vectors based on the first propensity score and the second propensity score, wherein the first machine learning model and the second machine learning model are retrained based on the updated state vectors.
2. The computer implemented method of claim 1 , wherein the first propensity score and the second propensity score are determined daily.
3. The computer implemented method of claim 1 , further comprising ranking messaging content based on the propensity score that exceeds the predetermined threshold to generate secondary outputs of at least one of the first machine learning model and the second machine learning model.
4. The computer implemented method of claim 3 , further comprising generating the message content to deliver to the user based in part on the ranked messaging content.
5. The computer implemented method of claim 1 , further comprising updating the current state data associated with the user based on the generated message content.
6. The computer implemented method of claim 5 , further comprising using the propensity score that exceeds the predetermined threshold to determine an effectiveness of the generated message content.
7. The computer implemented method of claim 1 , wherein the churn risk is either an indicator of whether the user will convert a trial membership to a subscriber membership or an indicator of whether the user will stop using the trial membership.
8. A system comprising:
a non-transitory data storage device; and
one or more special purpose computing devices that access and store data on the data storage device and employ at least one processor to perform actions, including:
accessing data of a user at a data source, the data of the user comprising historical data and current state data associated with the user;
parsing the accessed data of the user to identify event instances;
determining an initial propensity score based on the historical data and the current state data associated with the user;
in response to determining that the initial propensity score indicates a high churn risk associated with the user:
identifying a first hypothetical scenario associated with a first trained hypothetical model of a first machine learning model and a second hypothetical scenario associated with a second trained hypothetical model of a second machine learning model, wherein each of the first and the second hypothetical scenarios include a corresponding associated software service provider feature, among a plurality of software service provider features, a corresponding hypothetical attribute of the user, and a corresponding hypothetical state data of the user, wherein the first machine learning model and the second machine learning model were trained by segmenting a collection of historical data and current state data, stored at the data source as events and state vectors, into a training set, a test set, and a validation set and passing the training set, the test set, and the validation set through the first machine learning model and the second machine learning model;
implementing the first hypothetical model of the first machine learning model to determine a first propensity score associated with the first hypothetical scenario based on the corresponding hypothetical attribute and the corresponding hypothetical state data, wherein the first propensity score represents a higher propensity score gain
implementing the second hypothetical model of the second machine learning model to determine a second propensity score associated with the second hypothetical scenario based on the corresponding hypothetical attribute and the corresponding hypothetical state data, wherein the second propensity score represents a lower propensity score gain;
determining whether the first or the second propensity score meets a predetermined threshold;
generating message content to deliver to the user based on the determined propensity score that exceeds the predetermined threshold and on the at least one service provider feature; and
updating the state vectors based on the first propensity score and the second propensity score, wherein the first machine learning model and the second machine learning model are retrained based on the updated state vectors.
9. The system of claim 8 , wherein the first propensity score and the second propensity score are determined daily.
10. The system of claim 8 , further comprising ranking messaging content based on the propensity score that exceeds the predetermined threshold to generate secondary outputs of at least one of the first machine learning model and the second machine learning model.
11. The system of claim 10 , further comprising generating the message content to deliver to the user based in part on the ranked messaging content.
12. The system of claim 8 , further comprising updating the current state data associated with the user based on the generated message content.
13. The system of claim 12 , further comprising using the propensity score that exceeds the predetermined threshold to determine an effectiveness of the generated message content.
14. The system of claim 8 , wherein the initial propensity score is either an indicator of whether the user will convert a trial membership to a subscriber membership or an indicator of whether the user will churn.
15. A computing system comprising:
one or more processors; and
one or more non-transitory computer-readable storage devices storing computer-executable instructions, the instructions operable to cause the one or more processors to perform operations comprising:
accessing data of a user at a data source, the data source of the user comprising historical data and current state data associated with the user;
parsing the accessed data of the user to identify event instances;
determining an initial propensity score based on the historical data and the current state data associated with the user;
in response to determining that the initial propensity score indicates a high churn risk associated with the user:
identifying a first hypothetical scenario associated with a first trained hypothetical model of a first machine learning model and a second hypothetical scenario associated with a second trained hypothetical model of a second machine learning model, wherein each of the first and the second hypothetical scenarios include a corresponding associated software service provider feature, among a plurality of software service provider features, a corresponding hypothetical attribute of the user, and a corresponding hypothetical state data of the user, wherein the first machine learning model and the second machine learning model were trained by segmenting a collection of historical data and current state data, stored at the data source as events and state vectors, into a training set, a test set, and a validation set and passing the training set, the test set, and the validation set through the first machine learning model and the second machine learning model;
implementing the first hypothetical model of the first machine learning model to determine a first propensity score associated with the first hypothetical scenario based on the corresponding hypothetical attribute and the corresponding hypothetical state data, wherein the first propensity score represents a higher propensity score gain;
implementing the second hypothetical model of the second machine learning model to determine a second propensity score associated with the second hypothetical scenario based on the corresponding hypothetical attribute and the corresponding hypothetical state data, wherein the second propensity score represents a lower propensity score gain;
determining whether the first or the second propensity score meets a predetermined threshold;
generating message content to deliver to the user based on the determined propensity score that exceeds the predetermined threshold and on the at least one service provider feature; and
updating the state vectors based on the first propensity score and the second propensity score, wherein the first machine learning model and the second machine learning model are retrained based on the updated state vectors.
16. The computing system of claim 15 , wherein the first propensity score and the second propensity score are determined daily.
17. The computing system of claim 15 , further comprising ranking messaging content based on the propensity score that exceeds the predetermined threshold to generate secondary outputs of at least one of the first machine learning model and the second machine learning model.
18. The computing system of claim 17 , further comprising generating the message content to deliver to the user based in part on the ranked messaging content.
19. The computing system of claim 15 , further comprising updating the current state data associated with the user based on the generated message content.
20. The computing system of claim 19 , further comprising using the propensity score that exceeds the predetermined threshold to determine an effectiveness of the generated message content.